Application of the key characteristics of carcinogens in cancer hazard identification
Application of the key characteristics of carcinogens in cancer hazard identification
Kathryn Z.Guyto n 2
Ivan Rusyn 1
Weihsueh A.Chiu 1
Denis E.Corpet 0
Martin van den Ber g Emailm: 6
Matthew K.Ross 5
David C.Christian 4
Frederick A.Beland 8
Martyn T.Smit h 7
0 ,ENVT, INRA TOXALIM (Research Center in Food Toxicology), Université de Toulouse , Toulouse , France
1 cDee,partment of Veterinary Integrative Biosciences, Texas A&M University, College Station , TX 77843 , USA
2 Monographs Programme, International Agency for Research on Cancer , Lyon, Fran
3 Department of Medicine, Massachusetts General Hospital/Harvard Medical School , Boston, MA 02115 , USA
4 ,Department of Environmental Health, Harvard TH Chan School of Public Health , Boston, MA 02115 , USA
5 ,Center for Environmental Health Sciences, Department of Basic Sciences, College of Veterinary Medicine, Mississippi State University , Mississippi State, MS 39759 , USA
6 I,nstitute for Risk Assessment Sciences (IRAS), Utrecht University , Utrecht , The Netherlands
7 Division of Environmental Health Sciences, School of Public Health, University of California , Berkeley, CA 94720 , USA
8 Division of Biochemical Toxicology, National Center for Toxicological Research , Jefferson, AR 72079 , USA
Smith et al. (Env. Health Perspect. 124: 713, 2016) identified 10 key characteristics (KCs), one or more of which are commonly exhibited by established human carcinogens. The KCs reflect the properties of a cancer-causing agent, such as 'is genotoxic,' 'is immunosuppressive' or 'modulates receptor-mediated effects,' and are distinct from the hallmarks of cancer, which are the properties of tumors. To assess feasibility and limitations of applying the KCs to diverse agents, methods and results of mechanistic data evaluations were compiled from eight recent IARC Monograph meetings. A systematic search, screening and evaluation procedure identified a broad literature encompassing multiple KCs for most (12/16) IARC Group 1 or 2A carcinogens identified in these meetings. Five carcinogens are genotoxic and induce oxidative stress, of which pentachlorophenol, hydrazine and malathion also showed additional KCs. Four others, including welding fumes, are immunosuppressive. The overall evaluation was upgraded to Group 2A based on mechanistic data for only two agents, tetrabromobisphenol A and tetrachloroazobenzene. Both carcinogens modulate receptor-mediated effects in combination with other KCs. Fewer studies were identified for Group 2B or 3 agents, with the vast majority (17/18) showing only one or no KCs. Thus, an objective approach to identify and evaluate mechanistic studies pertinent to cancer revealed strong evidence for multiple KCs for most Group 1 or 2A carcinogens but also identified opportunities for improvement. Further development and mapping of toxicological and biomarker endpoints and pathways relevant to the KCs can advance the systematic search and evaluation of mechanistic data in carcinogen hazard identification. Introduction be carcinogens. The cancer hazard classification methodology, o-ut lined in the IARC Preamble 1(), is based on the systematic assemTheInternationalAgencyforResearchonCancer(IARC)Monographs bly, review and integration of evidence of cancer in humans, cancer Programme identifies the causes of human cancer. IARC assembles in experimental animals and cancer mechanisms. To date, more expert Working Groups, free of conflicts of interest, to evaluate than 1000 agents have been classified. IARC evaluations are used agents that have evidence of human exposure and are suspected to
Abbreviations many carcinogens evaluated prior to Volume 100, few data were
AhR aryl hydrocarbon receptor available on some mechanisms of recognized importance in c-ar
AOP adverse outcome pathway cinogenesis, such as epigenetic alterations17(). Fourth, the
evaluKCs key characteristics ation of mechanistic and other relevant data has been further
MOA mode of action challenged by the lack of a systematic and transparent method
NTP National Toxicology Program to search for and assemble mechanistic data for cancer hazard
PCP pentachlorophenol identification. Specifically, there was no widely accepted method
TCAB tetrachloroazobenzene to search systematically for relevant mechanisms, resulting in
TCDD 2,3,7,8-tetrachlorodibenzo-p-dioxin. a lack of uniformity in the mechanistic topics addressed across
assessments. Finally, there was no procedure to efficiently org-an
worldwide by national and international health agencies to supportize, analyze and interpret the voluminous mechanistic studies.
a wide range of subsequent activities ranging from research, to risk To address these challenges, the KCs of human carcinogens
assessment, to preventative actions2)(. were recently introduced as the basis of a uniform approach for
Of the around 120 agents classified in Group 1 (carcinogenic searching, organizing and evaluating mechanistic evidence to
to humans; seeTable 1) by IARC, most have ‘sufficient’ evidence support cancer hazard identification16(). The KCs comprise the
for their carcinogenicity in humans based on epidemiological properties of known human carcinogens. These characteristics
studies. However, epidemiological studies of cancer in exposed are distinct from the hallmarks of cancer that relate to the-prop
humans are often limited in number and may have deficiencies erties of cancer cells1(
) but instead reflect the ability of a
in terms of sample size, confounding and exposure characte-ri carcinogen to, for example, be genotoxic, be immunosuppressive
zation. Furthermore, for chemicals that have been introduced or modulate receptor-mediated effects (seTeable 2). Established
recently on the market, epidemiological studies may not exist or human carcinogens commonly exhibit one or more of these ch-ar
be relevant because of the long latency period for cancer de-vel acteristics, and, therefore, data on these characteristics can-pro
opment. Lifetime rodent cancer bioassays are declining in nu-m vide independent evidence of carcinogenicity when human data
ber, with only a fraction of the ~75 000 in the Toxic Substances are lacking. Such data can also help in interpreting the relevance
Control Act Chemical Substance Inventory having been formally and importance of findings of cancer in animals and in humans.
evaluated by the National Toxicology Program (NTP3),4(). In Herein, we describe and discuss the application of these
contrast, data on cancer mechanisms from human biomarker KCs in IARC Monograph evaluations that have taken adva-n
studies, in vivo animal tests andin vitro cell culture models are tage of the systematic consideration of mechanistic evidence.
increasing in both volume and diversity5(–8). This includes The strengths and weaknesses of this approach are discussed,
human in vitro test systems to detect effects on a variety of b-io as are opportunities for further progress and refinement. We
logical receptors and to explore genetic susceptibili6t,y9)(. further discuss how the paradigm could be expanded to other
When the evidence from human epidemiologic studies is less endpoints and how future toxicological and molecular ep-ide
than sufficient, strong mechanistic data can play a pivotal role in miological studies could be developed to generate more useful
the overall carcinogen hazard classificatio1n0,(
). For instance, information for the process of carcinogen evaluation.
even though the evidence from rodent cancer bioassays provided
‘sufficient’ evidence of cancer in animalsd,-limonene was class-i
fied in Group 3 (not classifiable as to its carcinogenicity to humans) Methods
on the basis of mechanistic and other relevant data because the Methods and results of mechanistic evaluations were compiled from the
probable mechanism of carcinogenicity in animals was unlikely more than 30 agents evaluated in Meetings 112–119 of the IARC Monographs
to be operative in humans 1(2). Other agents have been classified (
), based on the Lancet Oncology publications21(–28) and published
in Group 2A (probably carcinogenic to humans) or even in Group 1 Monographs (29). As noted in the Preamble to the IARC Monographs1() and
(carcinogenic in humans) based on strong evidence for reco-g Ionrsgtarnuicztaitoinons taondAuevtahlourast3(i(0o)n;spereocaeldsour(3e1s)w), etrheeamsfeoclhlaonwiss:tic data search,
nized carcinogen mechanisms, e.g. genotoxicity (ethylene oxide
)), inhibiting DNA repair (etoposide 1(
)) or binding to the aryl • A working list of endpoints associated with each KC was developed by
hydrocarbon receptor (AhR) and subsequent downstream effects IARC in collaboration with the expert Working GroupSsu(pplementary
)). Table 1, available atCarcinogenesis Online).
A recent review of these agents and all other Group 1 human • For the Monograph Meetings 112–1192(0,29), IARC Monographs staff
carcinogens classified in Volumes 1–99 identified several issues pweorrfkoirnmgeldisitnoitfiaselalricthertaetrumres fsoerartchheeKsCfosrdsetvueldoiepsedonbytIhAeRCagiennctoullsai-nbgo a
relevant to improving the evaluation of mechanistic data for-car ration with the expert Working Groups (seSeupplementary Table ,2
cinogen hazard identification1(
). First, it was noted that many available atCarcinogenesis Online). Additional complementary lite-ra
human carcinogens showed a number of characteristics that are ture searches by IARC and the Working Groups incorporated terms for
shared among carcinogenic agents. Second, different human c-ar the agent and metabolites, alone or in combination with terms related
cinogens may exhibit a different spectrum of these key charact-er to carcinogenicity. Literature identified through other searches (e.g.
istics (KCs) and operate through distinct mechanisms. Third, for from reference lists of retrieved articles, past Monographs or other
authoritative reviews) was also included.
• The literature review, study exclusion and categorization of included
Table 1. Definitions of the IARC classifications studies were performed by the Volume 112 expert Working Group
Classification Definition members and by IARC Monographs staff for subsequent meetings,
with further review and refinement by each expert Working Group.
Group 1 Carcinogenic to humans Specifically, the steps involved entailed:
Group 2A Probably carcinogenic to humans o Based on title and abstract review, studies were excluded if they
Group 2B Possibly carcinogenic to humans were not about the agent (or a metabolite) or if they reported no
Group 3 Not classifiable as to its carcinogenicity to humans data on mechanistic or toxicological endpoints.
Group 4 Probably not carcinogenic to humans ○ Itnhcelu10deKdCsst(usdeeieTsawbleer e2),fubratsheedrosnorttheedminetcohacnaitsetgicoreinesdproepinrtesseanntding
Each IARC Working Group considered the extent to which relevant me-ch
anistic events had been established, including whether the mechanistic
event has been challenged experimentally, as well as the consistency
of the results in different experimental systems and the overall co-her
ence of the database 1(). As a result, the evidence was classified based
on collective expert judgment as ‘strong, moderate or weak’ based on very hot beverages 2(0,29). Of note, this particular set of agents
the assembled mechanistic information. As outlined in the Instructions may not be representative of all carcinogens, as a much broader
to Authors (30), these classifications were based on the extent of data set of complex mixtures, occupational exposures, physical
available that met the criteria in the IARC Monographs Preamb1l)e, ( agents, biological agents and lifestyle factors have been cl-assi
addressed the range of study designs and doses/concentrations tested, fied in Group 1, as recently reviewed in Volume 100 of the IARC
and included consideration of whether or not the effects were observed Monographs (29).
at the physiological, cellular or molecular level, as well as the presence
of any consistencies or differences in results within and across expe-ri
mental designs. Emphasis was given to existing mechanistic data from Literature search results
human-related studies, and gaps in evidence were identified. In general, The extent of available evidence on cancer mechanisms was
in experimental systems, studies of repeated doses and of chronic exp-o variable across the range of agents evaluateTda(ble 3). Agents
sures were accorded greater importance than studies of a single dose or classified in Group 1 had generally been the subject of more
time point. Consideration was also given to the extent of concurrent-tox extensive mechanistic studies than agents classified in Groups
icity that was observed. Route of exposure was considered a less imp-or 2 or 3. For example, as illustrated iFnigures 1 and 2, 239 articles
ttahnetexfapcotsourreins athned etvaarlgueatttioisnsoufesexmpearyi mvaernytaalcrsotsusdeiexsp,einrimreecnotganlitmioondtehlsat on KCs were identified for pentachlorophenol (PCP), compared
and in exposed human populations. Another important aspect of the with 35 for 2,4,6-trichlorophenol, despite the similar use scen-ar
evaluation was the assessment of whether the mechanism can operate ios of these two agents. For most agents evaluated as Group 2B
in humans (1). Accordingly, human biomarker studies incorporating en-d or 3, data on KCs were sparse. For instance, for more than half
points relevant to the KCs of carcinogens were deemed to be especially of the 16 agents evaluated in Group 2B, fewer than 30 articles
valuable when available. were included on all KCs following targeted literature searches
(e.g. for 1-bromopropane, 3-chloro-2-methylpropene, furfuryl However, strong evidence for two of these KCs (is genotoxic,
alcohol, indium tin oxide, 1t-ert-butoxypropanol, molybdenum induces oxidative stress) was found for malathion, diazinon and
trioxide, N,N-dimethyl-p-toluidine, tetrachlorvinphos and t-et glyphosate, with malathion additionally showing three others
rahydrofuran). A notable exception was 2,4-dichlorophenox-y (induces chronic inflammation, modulates receptor-mediated
acetic acid (2,4-D), for which more than 200 studies on KCs effects and alters cell proliferation21),(31).
were identified. However, nearly half of these publications c-on Other Group 1 and Group 2A agents showed strong evidence
cerned the genotoxicity of 2,4-D, a KC for which there was not for distinct sets of KCs. For instance, welding fumes induce
strong evidence. Of agents evaluated in Group 2B or 3, coffee chronic inflammation and are immunosuppressive27(). The
was among the most widely studied, with over 300 publications strong evidence encompassed numerous panel studies of
crosscovering a broad range of KCs. As part of the large database of sectional and cohort design demonstrating increases in lung
human observational and experimental studies, these mech-a and systemic inflammation biomarkers. Risk for infection in
nistic studies were supportive but not influential in the o-ver exposed workers was increased in epidemiological studies of
all evaluation (which was based on ‘inadequate’ evidence in different design. The consistent and coherent evidence from
). For all Group 1 and most Group 2A agents [with human studies of welders was further supported by findings in
the exception of 2-mercaptobenzothiazole, tetrachloroazob-en experimental systems showing that welding fumes impaired
zene (TCAB) and very hot beverages], a substantial evidence resolution of pulmonary infection and induced chronic p-ul
database on the mechanisms of carcinogenicity was identified, monary inflammation. Similar to welding fumes, chronic st-ud
encompassing hundreds of scientific publications and generally ies in rodents provided strong evidence for other agents that
covering a range of different KCs, endpoints and test systems induce chronic inflammation (e.g. malathion, TCAB, indium tin
utilizing various exposure routes and durations. oxide and melamine (
). On the other hand, there was
strong evidence for a different set of KCs (modulates
receptorKCs of carcinogens with strong evidence mediated effects, is immunosuppressive and induces oxidative
Several of the Group 1 and Group 2A agents showed strong e-vi stress) for dichlorodiphenyltrichloroethane (DDT) and tetra-bro
dence for common sets of KCs (Table 3). PCP and hydrazine are mobisphenol A (
metabolically activated to electrophilic moieties, are genotoxic, For the agents evaluated in Group 2B or Group 3, there was
induce oxidative stress and altered cell proliferation or death usually strong evidence for at most one KC (seTeable 3). An
(PCP) or nutrient supply (hydrazine)2(
). PCP additionally exception was 1-bromopropane, which had strong evidence of
modulates receptor-mediated effectsN. ,N-Dimethylformamide multiple KCs, primarily from experimental system2s4)(. This
(DMF) shared some of these KCs (is metabolically activated, strong evidence from experimental systems did not warrant a
induces oxidative stress and alters cell proliferation) but was change in the overall classification of 1-bromopropane, however,
not considered to have strong evidence for being genotoxic24(). as the IARC Monographs Preamble requires ‘strong evidence
from exposed humans’ (for Group 1) or that the mechanism has evidence that they induce oxidative stress adding support to
been shown to ‘also operate in humans’ (for Group 2A)1(). The findings for these other KCs (e.g. oxidatively damaged DNA).
agents with strong evidence for one KC included furfuryl alcohol, However, oxidative stress is not unique to cancer, and strong
which is metabolically activated to the electrophilic 2-sulph-oxy evidence that 2,4-D induces oxidative stress did not result
methylfuran, while 2,4-D induces oxidative stress and indium in a change in the overall cancer hazard classification. Other
tin oxide induces chronic inflammation2(2,27,28). Coffee drink- KCs seen with multiple agents were ‘is genotoxic’ (six agents),
ing was associated with antioxidant effects25(). Only one agent ‘induces chronic inflammation’ (six agents), ‘alters cell proli-fera
(3-chloro-2-methylpropene) was determined to be genotoxic tion, cell death or nutrient supply’ (six agents), ‘is electrophilic
(24), although data for this KC were broadly available for many or can be metabolically activated’ (five agents), ‘is immunos-up
compounds under consideration, usually from the standard pressive’ (five agents) and ‘modulates receptor-mediated effects’
battery of genotoxicity assays performed in conjunction with (five agents). No agent had strong evidence for three KCs (alters
chronic bioassays and from published reports. Similarly, some DNA repair or causes genomic instability, induces epigenetic
agents (e.g. melamine,β-myrcene) were specifically noted as not alterations and causes immortalization).
being genotoxic (28); a lack of genotoxicity is one of the seven
criteria that must all be met in order to conclude that rodentImpact of mechanistic data on the overall
carcinogens operate through a mechanism not relevant to evaluation of carcinogenicity by IARC Monograph
humans (i.e. due to precipitates in the bladder oαr2u-globulin Working Groups
in the kidney) (34). Regarding the latter, four of the agents eva-lu The agents with ‘sufficient’ or ‘limited’ evidence of cancer in
ated (1-tert-butoxypropanol,β-myrcene, pyridine and tetrah-y humans often exhibited strong evidence of multiple KCs; ho-w
drofuran) increased kidneyα2u-globulin in male rats, but none ever, the mechanistic data did not affect the overall classi-fica
satisfied the seven IARC criteria for anα2u-globulin-associated tion of these agents. The classification of four agents in Group 1
tumorigenic response 2(
). (e.g. PCP, lindane, welding fumes and consumption of processed
The KC most often showing ‘strong’ evidence across all 34 meat) was based on ‘sufficient’ evidence of cancer in humans
agents evaluated in IARC Monograph Meetings 112–119 was (
). When available, ‘limited’ evidence of cancer in
‘induces oxidative stress’ (11 agents; seeFigure 3). For agents humans was the basis of Group 2A evaluations, alone (diazinon
that induce oxidative stress (e.g. PCP, 2,4-D2(
)), experimen- and consumption of red meat) or together with ‘sufficient’ -evi
tal studies of the effect of antioxidants or in knockout animals dence of cancer in experimental animals (malathion, hydrazine,
were part of the strong dataset. For all but one agent (2,4-D), this DDT, DMF, glyphosate, 2-mercaptobenzothiazole, very hot b-ev
KC was accompanied by strong evidence for one or more other erages, dieldrin and aldrin metabolized to dieldrin2)1–(
KCs (see Table 3). This is perhaps not surprising, given the close mechanistic data supported a change in the overall evaluation
association of the underlying cancer mechanisms, with strong for very few agents, only two, tetrabromobisphenol A and TCAB,
an e e e e an
iercannC iiffetcSun iiffetcSun iiffetcSun Ietaadqun iiffetcSun iiffetcSun iiffetcSun iiffetcSun iifftceSnu iifftceSnu iietLdm iiffetcSun iiffetcSun Ietaadqun iiffetcSun iiffetcSun iiffetcSun iietLdm iiffetcSun iiffetcSun iiffetcSun iiffetcSun iiffetcSun iiffetcSun iiffetcSun iiffetcSun iiffetcSun iiffetcSun iiffetcSun iiffetcSun iiffetcSun iiffetcSun Ietaadqun Ietaadqun .r2oApuG sertcoghnm
X X X X
K X X X X X X X X X
lo lo e id
n an en ixd lou s ed )t iev lpp oh
etAgn PPC ieLadnn ifleegdunmW istoopunCnm iltoaahnM ierzyadnH TDD -FDM,NN BCTA itrrseeoaobbphTm iizoannD lsetyoaphG -eert2cboapM istoopunCnm iileraddnnD iilertoddn eertyvbohV -rr1oooBpm ,i-lrc42oohD ,)(-42D ---lrt32yoohhCm ()ileertccagdhn flllrrFcyooauuh iiiIetxoddunnm ileeanMm --rt1yxooBpputtre ,,i-lrr642coophhT β-eercynM ilertyboodunMm i---lettyhDm,pNN irtPoaahn iieryPdn ilerrtcvooaphhTn ferrrtyoaaduhTn iiilleercyodhnnV iiffeerkogdCnn (eerttvyooahnM ssseeerrrttogpnnX aifsseertboaduuNmbilseettoAaappdnmc.iiffseetttcxoAadnn
of the 34 evaluations in Monograph Meetings 112–119. In both In a research setting, it is essential that scientific hypotheses
cases, the mechanistic data provided strong evidence of m-ul can be continually modified, updated and repeatedly tested.
tiple KCs and supported an upgrade in the cancer hazard cl-as However, in the context of hazard evaluation and risk asse-ss
sification, from Group 2B to Group 2A2(
). ment, a hypothesis-based paradigm can introduce bias because
Specifically, tetrabromobisphenol A was classified in focusing on a specific set of hypothesized key events inherently
Group 2A on the basis of sufficient evidence of carcinogenicity results in exclusion of data, leading to analyses that favor one
in animals and the strong mechanistic evidence on three KCs or more particular mechanisms or series of key events. As a
(modulates receptor-mediated effects, is immunosuppressive related challenge, an uneven level of experimental results can
and induces oxidative stress) that were shown to ‘also op-er result from disproportionate resources having been focused on
ate in humans’ (
). On the other hand, TCAB was classified investigating a favored mechanism1(). Moreover, as biological
in Group 2A because it belongs, on the basis of mechanistic knowledge develops, hypothesized key events and MOA/AOP
considerations, to the class of agents that activate AhR, and may be shown to be incorrect or incomplete3,(41).
some members of this class have previously been evaluated as Other efforts to establish systematic approaches for revi-ew
Group 1 or Group 2A carcinogens1(
). In addition to the strong ing mechanistic studies have been initiated, such as by gat-h
evidence of multiple KCs, TCAB has structural resemblance to ering information on postulated mechanisms (e.g. blood IGF1
AhR agonists classified in Group 1, causing a similar pattern of and prostate cancer) 4(2,43). A separate approach sought to
tumors in experimental animals and induces pathognomonic broaden consideration of cancer mechanisms by organizing
responses for AhR activation (e.g. chloracne in multiple species). them according to the hallmarks of cance4r4)(. However, as
the hallmarks characterize frank tumors that are evident at the
end of the carcinogenic process 1(
), they are distinct from
Discussion the effects of carcinogens and may not capture fully important
Mechanistic data are an integral part of all cancer hazard e-valu mechanistic effects that occur early or transiently throughout
ations and by extension of all human health risk assessment tumor development.
) but still pose a challenge to decisionmakers as Here, we have described our recent experience with a
there is no generally accepted procedure to efficiently org-an novel alternative approach that does not require postulating
ize, analyze and interpret the voluminous number of mech-a the mechanism or MOA by which a chemical causes cancer in
nistic studies. One framework that has been widely used and animals and/or humans. Rather, the approach utilizes the KCs
promoted is the mode of action (MOA) approach 3(
), which of human carcinogens as the basis of a uniform approach to
subdivides the pathway between an agent (exposure) and an search, organize and evaluate mechanistic evidence to support
adverse effect into a series of hypothesized key events. This cancer hazard identification1(
). These KCs comprise 10
properpathway-based approach has been further developed into the ties of known human carcinogens. The 2017 National Academy
more recent adverse outcome pathway (AOP) framework, which of Sciences report on ‘Using 21st Century Science to Improve
was initially described by the ecotoxicology communit3y9)(. Risk-Related Evaluations’ recently opined that the KCs approach
However, the hypothesized key events that form the common ‘avoids a narrow focus on specific pathways and hypotheses
basis of MOAs and AOPs are inherently limited by the current and provides for a broad, holistic consideration of the mec-ha
understanding of the disease process. This limitation was r-ec nistic evidence.’ (45) The KCs have been used by the IARC
ognized by Sir Bradford Hill, who noted that ‘what is biologically Monographs Programme to evaluate the mechanistic data c-om
plausible depends upon the biological knowledge of the day’ piled for the more than 30 agents evaluated in Meetings 112–119
and described the biological plausibility of a suspected cau-sa of the IARC Monographs (
). As also indicated in the 2017
tion as ‘helpful’ but not necessary40( ). National Academy of Sciences report, the KCs approach could
be expanded to other endpoints, including endocrine disruption little overall impact on the mechanistic evaluation of an agent,
and cardiovascular disease4(
). This effort could be informed by analyses of the ToxCast/Tox21 assays revealed adequate co-v
the experience described herein, and the strengths and wea-k erage for very few of the KCs 1(1), a critical data gap that also
nesses of the approach highlighted below. merits consideration as these assays are applied in evaluation
Recent experiences with the KCs of carcinogens reveal a of other toxicological endpoints.
number of strengths, such as allowing for a systematic search Opportunities for refinement that are also relevant to other
and assembly of the relevant literature, followed by organ-iza applications of the approach include (i) clearly delineating the
tion of the studies into easily reviewable groupings (seFeigures toxicological and biomarker endpoints associated with each KC;
1 and 2). This review process revealed strong evidence of KCs (ii) developing sensitive and specific literature search terms for
for many agents classified in the higher categories of Group 1 each KC and applying the latest bioinformatic tools in literature
or 2A (see Figure 3). Indeed, 11 of 12 agents with strong ev-i searching; (iii) promoting uniformity of evaluations through
dence for two or more KCs were classified in Group 1 or 2A. On the documentation and clarification of procedures by the IARC
the other hand, for five other classifications in Group 1 or 2A Secretariat, including development of examples of ‘strong’ -evi
(dieldrin and aldrin metabolized to dieldrin, consumption of dence for a KC based on the extent of testing in different s-ys
processed meat, consumption of red meat, very hot beverages tems (in humans, in chronic andin vivo settings in rodents and
and 2-mercaptobenzothiazole), there was no strong evidence in vitro); (iv) exploring approaches to integrate evidence across
for any KC. For dieldrin, a substantial literature was identified, KCs (e.g. when evidence of one KC would be alone sufficient to
but coverage of the KCs was incomplete, and findings lacked result in any change in the overall cancer hazard classification
consistency and coherence. Many mechanistic studies were and whether evidence of specific groups of KCs would otherwise
also identified for meat, but studies of the consumption of the be needed) and (v) soliciting feedback after each Monograph
meat itself were sparse, even though important components meeting to continue to identify issues arising in the evaluation
of red and processed meat showed strong evidence for ce-r and opportunities for progress.
tain KCs. For very hot beverages and 2-mercaptobenzoth-ia Overall, we show that application of the KCs to hazard id-en
zole, only a few mechanistic studies were available. Very hot tification is a robust new approach to organizing and eval-uat
beverages are difficult to study mechanistically, even though ing the mechanistic data related to carcinogenesis that avoids
heat is obviously cytotoxic and will induce tissue damage the need to identify, and thereby restrict attention to, specific
and inflammation. Finally, for most agents (with welding as pathways and hypotheses. Because they are based on empir-i
a notable exception 2(
)), few studies of biomarker endpoints cal observations of characteristics associated with known -car
relevant to the KCs in exposed humans were available. Such cinogens, the KCs thus provide a more agnostic and unbiased
studies have been encouraged by the National Academy of survey of the mechanistic literature. This approach has been
Science of the USA (
), and future effort in this area will be successfully applied in eight IARC Monograph meetings eva-lu
of importance and relevance to any evaluation of evidence ating 34 agents, but this experience also points to a number of
for KCs. potential refinements that can further improve the evaluation
For two agents, tetrabromobisphenol A and TCAB, out of mechanistic data to support identification of human carc-ino
of 34 total, strong evidence for distinct KCs resulted in their genic agents.
upgrade to Group 2A even though the epidemiological findings
were considered inadequate (
). Oxidative stress was the Supplementary material
tomhtohesetargcKeoCnmstsmineovanalllKucCaatsieeddse,nebxtucifteipetdthfiwosriwt2ha,4ss-tDfor.uo2nn,4dg-Dienvwicdaoesnncjculeanscastcifiroioensdswaasitlahl Sounplipnlee.mentary Tables 1 and 2 can be found atCarcinogenesis
Group 2B carcinogen 2(2), and a subsequent meta-analysis has
provided new evidence for an association between lymphoma Funding
and exposure to 2,4-D (46). However, noncarcinogens can also We gratefully acknowledge financial support from the National
induce oxidative stress, and so, this KC should be interpreted Institutes of Health, USA (U01 CA33193, K.Z.G.; Superfund grant
with caution. Overall, a structured approach based on KCs NIH P42ES004705, M.T.S.; Superfund grant P42ES027704, I.R. and
performed well in helping evaluate the carcinogenicity of the W.A.C.) and the European Union Programme for Employment
34 agents discussed here, and the agents with the strongest and Social Innovation “EaSI” (2014-2020) (K.Z.G.).
evidence of carcinogenicity had more KCs with strong as a
descriptor.In all cases, the use of a transparent search tool and
the binning of data clearly helped the expert working group Acknowledgements
organize and evaluate the mechanistic data and is an imp-or The authors gratefully acknowledge the IARC Working Groups,
tant element to take forward in the extension of the approach IARC secretariat and other participants for IARC Monograph
to other endpoints. Meetings 112–119. The views expressed in this manuscript
Several weaknesses, however, could be addressed moving do not necessarily represent those of the U.S. Food and Drug
forward. The mechanistic data in general, and especially novel Administration or the European Commission.
high-throughput data stream from very large toxicity -test Conflict of Interest Statement: None of the authors have any conflict
ing programs, are not without important limitations. These of interest regarding the chemical agents or topics discussed.
have been extensively covered elsewhere 4(
), so only a
few points will be highlighted. The major technical limitations
include general deficiency in metabolic capacity in assays, l-im References
ited biological coverage of the mechanistic assays for the KCs 1. IARC. (2006) Preamble to the IARC Monographs. http://monographs.iarc.
), extrapolation oifn vitro to in vivo exposures and challenges fr/ENG/Preamble/index.php.
with validation, which cannot match the pace of development 2. Pearce, N.et al. (2015) IARC monographs: 40 years of evaluating carcin-o
of new assays, models and test systems. In addition to having genic hazards to humans. Environ. Health Perspect., 123, 507–514.
l d g o e a B r r ;e o e c t ev n s b ed ie t m uNm tSdu (llAK 23 37 18 14 24 11 74 17 2 14 12 14 1 14 23 3 2 26 1 2 2 7 3 3 1 20 4 2 2 7 30 3 ec 9 5 9 a4 9 7 5 0 2 7 5 6 9 a4 7 b0 9 9 1 1 2 8 2 5 4 9 9 9 5 9 2 6 4 b0 fro n e d i s v l e a
3. Guyton , K.Z. et al. ( 2009 ) Improving prediction of chemical carcin-o 27 . Guha , N. et al.; International Agency for Research on Cancer Monograph genicity by considering multiple mechanisms and applying toxic-og Working Group . ( 2017 ) Carcinogenicity of welding, molybdenum tri-ox enomic approaches . Mutat. Res. , 681 , 230 - 240 . ide, and indium tin oxide . Lancet. Oncol. , 18 , 581 - 582 .
4. National Toxicology Program. (2017T)esting Information . https://ntp. 28. Grosse , Y. et al.; International Agency for Research on Cancer niehs .nih.gov/testing/index.htm.l Monograph Working Group. ( 2017 ) Some chemicals that cause tumours
5. National Academy of Science. ( 2007 ) Applications of Toxicogenomic of the urinary tract in rodents . Lancet. Oncol., 18 , 1003 - 1004 . Technologies to Predictive Toxicology and Risk Assessment . National 29. IARC . ( 2017 ) Monographs and Supplements Available Online. http://monoAcademies Press, Washington, DC. graphs.iarc.fr/ENG/Monographs/PDFs/index.ph.p
6. Fielden , M.R. et al. ( 2018 ) Modernizing human cancer risk assessment 30 . IARC. ( 2017 ) Instructions to Authors for the Preparation of Drafts for IARC of therapeutics . Trends Pharmacol. Sci. , 39 , 232 - 247 . Monographs. http://monographs.iarc.fr/ENG/Preamble/instructions.
7. Tice , R.R. et al. ( 2013 ) Improving the human hazard characterization of php. chemicals: a Tox21 update . Environ. Health Perspect. , 121 , 756 - 765 . 31. IARC Working Group on the Evaluation of Carcinogenic Risks to
8. Collins , F.S. et al. ( 2008 ) Toxicology . Transforming environmental health Humans . ( 2017 ) Some Organophosphate Insecticides and Herbicides . protection. Science , 319 , 906 - 907 . International Agency for Research on Cancer, Lyon, France.
9. Zeise , L. et al. ( 2013 ) Addressing human variability in next-generation 32 . Straif , K. et al. ( 2014 ) Future priorities for the IARC monographs. Lancet. human health risk assessments of environmental chemicals . Environ. Oncol. , 15 , 683 - 684 . Health Perspect., 121 , 23 - 31 . 33 . Guha , N. et al. ( 2016 ) Prioritizing chemicals for risk assessment using
10. Cogliano , V.J. et al. ( 2008 ) Use of mechanistic data in IARC evaluations. chemoinformatics: examples from the IARC monographs on pes-ti Environ . Mol. Mutagen ., 49 , 100 - 109 . cides. Environ. Health Perspect., 124 , 1823 - 1829 .
11. Chiu , W.A. et al. ( 2018 ) Use of high-throughput in vitro toxicity scre-en 34 . IARC ( 1999 ) Species differences in thyroid, kidney and urinary bladder ing data in cancer hazard evaluations by IARC Monograph Working carcinogenesis . Proceedings of a Consensus Conference . Lyon, France, Groups. ALTEX, 35 , 51 - 64 . 3 - 7 November 1997 .
12. IARC Working Group on the Evaluation of Carcinogenic Risks to 35 . Berggren , E. et al. ( 2015 ) Chemical safety assessment using readHumans. (1999) Some Chemicals that Cause Tumours of the Kidney or across: assessing the use of novel testing methods to strengthen the Urinary Bladder in Rodents and Some Other Substances . International evidence base for decision making . Environ. Health Perspect., 123 , Agency for Research on Cancer, Lyon, France. 1232 - 1240 .
13. IARC Working Group on the Evaluation of Carcinogenic Risks to 36 . Sturla , S.J. et al. ( 2014 ) Systems toxicology: from basic research to risk Humans. ( 1994 ) Some Industrial Chemicals. International Agency for assessment . Chem. Res. Toxicol. , 27 , 314 - 329 . Research on Cancer, Lyon, France. 37. Birnbaum , L.S. et al. ( 2016 ) Informing 21st-century risk assessments
14. IARC Working Group on the Evaluation of Carcinogenic Risks to with 21st-century science . Environ. Health Perspect. , 124 , A60 - A63 . Humans. ( 2012 ) Pharmaceuticals . International Agency for Research 38. Meek , M.E. et al. ( 2014 ) New developments in the evolution and applic-a on Cancer, Lyon, France. tion of the WHO/IPCS framework on mode of action/species concor-d
15. IARC Working Group on the Evaluation of Carcinogenic Risks ance analysis . J. Appl. Toxicol. , 34 , 1 - 18 . to Humans. ( 1997 ) Polychlorinated Dibenzo-para-dioxins and 39 . Ankley , G.T. et al. ( 2010 ) Adverse outcome pathways: a conceptual Polychlorinated Dibenzofurans. International Agency for Research on framework to support ecotoxicology research and risk assessment . Cancer , Lyon, France. Environ. Toxicol. Chem., 29 , 730 - 741 .
16. Smith , M.T. et al. ( 2016 ) Key characteristics of carcinogens as a basis 40 . Hill , A.B. ( 1965 ) The environment and disease: association or caus-a for organizing data on mechanisms of carcinogenesis . Environ. Health tion? Proc. R. Soc. Med ., 58 , 295 - 300 . Perspect., 124 , 713 - 721 . 41 . Rusyn , I. et al. ( 2012 ) Mechanistic considerations for human relevance
17. Herceg , Z. et al. ( 2013 ) Towards incorporating epigenetic mechanisms into of cancer hazard of di(2-ethylhexyl) phthalate . Mutat. Res. , 750 , carcinogen identification and evaluation . Carcinogenesis , 34 , 1955 - 1967 . 141 - 158 .
18. Hanahan , D. et al. ( 2011 ) Hallmarks of cancer: the next generation . Cell , 42 . Harrison , S. et al. ( 2017 ) The albatross plot: a novel graphical tool for 144, 646 - 674 . presenting results of diversely reported studies in a systematic review .
19. Hanahan , D. et al. ( 2000 ) The hallmarks of cancer . Cell , 100 , 57 - 70 . Res. Synth. Methods , 8 , 281 - 289 .
20. IARC. Past Meetings-Recently Evaluated. http://monographs.iarc.fr/ENG/ 43. Harrison, S. et al. ( 2017 ) Does milk intake promote prostate cancer i-ni Meetings/index1.php. tiation or progression via effects on insulin-like growth factors (IGFs)?
21. Guyton , K.Z. et al.; International Agency for Research on Cancer A systematic review and meta-analysis . Cancer Causes Control , 28 , Monograph Working Group, IARC , Lyon, France. ( 2015 ) Carcinogenicity 497-528 . of tetrachlorvinphos, parathion, malathion, diazinon, and glyphosate. 44. Goodson , W.H. III et al. ( 2015 ) Assessing the carcinogenic potential of Lancet . Oncol., 16 , 490 - 491 . low-dose exposures to chemical mixtures in the environment: the
22. Loomis , D. et al.; International Agency for Research on Cancer challenge ahead. Carcinogenesis , 36 , S254 - S296 . Monograph Working Group, IARC , Lyon, France. ( 2015 ) Carcinogenicity 45 . National Academy of Science ( 2017 ) Using 21st Century Science to of lindane, DDT, and 2,4-dichlorophenoxyacetic acid . Lancet . Oncol., Improve Risk-Related Evaluations . The National Academies Press, 16 , 891 - 892 . Washington, DC.
23. Bouvard , V. et al.; International Agency for Research on Cancer 46 . Smith , A.M. et al. ( 2017 ) 2 , 4 -Dichlorophenoxyacetic acid ( 2 , 4 - D ) and risk Monograph Working Group. ( 2015 ) Carcinogenicity of consumption of of non-Hodgkin lymphoma: a meta-analysis accounting for exposure red and processed meat . Lancet. Oncol. , 16 , 1599 - 1600 . levels. Ann. Epidemiol., 27 , 281 - 289 . e4 .
24. Grosse , Y. et al.; International Agency for Research on Cancer 47 . Kleinstreuer , N.C. et al. ( 2014 ) Phenotypic screening of the ToxCast Monograph Working Group . ( 2016 ) Carcinogenicity of some industrial chemical library to classify toxic and therapeutic mechanisms . Nat. chemicals. Lancet. Oncol. , 17 , 419 - 420 . Biotechnol., 32 , 583 - 591 .
25. Loomis , D. et al.; International Agency for Research on Cancer 48 . Kavlock , R. et al. ( 2012 ) Update on EPA's ToxCast program : providing Monograph Working Group. ( 2016 ) Carcinogenicity of drinking coffee, high throughput decision support tools for chemical risk man-age mate, and very hot beverages . Lancet. Oncol. , 17 , 877 - 878 . ment. Chem. Res. Toxicol. , 25 , 1287 - 1302 .
26. Guyton , K.Z. et al.; International Agency for Research on Cancer 49. National Academy of Science ( 2007 ) Toxicity Testing in the 21st Monograph Working Group . ( 2016 ) Carcinogenicity of pentachlorop-he Century: A Vision and A Strategy. National Academies Press, nol and some related compounds . Lancet. Oncol. , 17 , 1637 - 1638 . Washington, DC.